Benchmarking short and long read sequencing technologies using HG002 reference sample

Long read sequencing technologies, generating reads tens to hundreds kilobases in length, are becoming increasingly competitive and complementary to short read technologies in human whole genome sequencing. Using the HG002 reference sample, we benchmarked the variant calling performance of recent sequencing technologies, including Illumina short reads, Illumina Complete Long Reads (ICLR), Pacific Biosciences (PacBio), and Oxford Nanopore Technologies (ONT), under different depth-of-coverage conditions. We evaluated these technologies against the Genome in a Bottle Telomere-to-Telomere truth set, the Human Genome Structural Variation Consortium truth set, and the Challenging Medically Relevant Genes truth set. For single nucleotide variant detection, all technologies performed similarly at coverage depths between 12× and 30×. For structural variant detection, we observed major improvements in accuracy and sensitivity with long read approaches compared to short reads. Structural variant calling performance was stable at coverage depths between 15× and 30×. For indel detection, performance declined rapidly below 30× for all technologies. Our results confirmed that indel calling with ONT is less accurate than for other technologies in spite of the latest chemistry. The hybrid ICLR approach showed intermediate performance between Illumina short reads and long read technologies (PacBio and ONT) for both indels and structural variants. Long reads enhanced haplotype phasing resolution, enabling the phasing of over 80\% of the genome. Overall, these findings provide valuable insights into the different pros and cons of recent sequencing technologies, aiding the decision-making in future research projects, technological platforms development, and clinical applications.